From 10% to 100%: Infill Compression Strength

[@Haapavuo]

For all non-US: 0.225 pounds = 0.1 kg 10%: 0.2 kN = 20.4 kgf 20%: 0.5 kN = 51 kgf 40%: 1.7 kN = 173 kgf 50%: 2.5 kN = 255 kgf 60%: 3.4 kN = 347 kgf

[@M2rsh]

who uses kgf

[@Haapavuo]

@@M2rsh if you want to know how many kgs you can put on your print, kgf is way to go.

[@M2rsh]

@@Haapavuo It's literally N/10 (Or whatever g approximation you want to use) kgf is deprecated

[@Corruptedpuppy]

thank you😃

[@kriziaktv]

European here Newton is in kilos so whats kgf and why are you converting it since Newton is kg.m.s-2? And 10N is the gravitational force of 1kg so why kgf?

[@joshsekel]

It would be cool to see the difference a more isotropic infill like gyroid would handle this test. Also, a static infill and varying perimeter counts. Then just to round things out, different print temperature (maybe just outside normal temps), and then finally, different cooling levels. You didn't have anything better to do, right?

[@zerker2000]

Gradient infill, also :)

[@marcus3d]

Also include the weights of the printed parts, so we can do a weight/force graph to identify sweet spots.

[@lazyman1011]

Also change the direction of force for the same probe.

[@marcus3d]

@@lazyman1011 Yes, test both across layer lines and along layer lines.

[@dtibor5903]

Do it yourself 🤣

[@stevewells20]

I've done a lot of destructive testing on 3d printed parts, and I highly recommend making a centering jig for the part under test if you want good data. I had really inconsistent results until I put one on (the press side, not the passive side).

[@802Garage]

They plan on getting real lab equipment and making standardized test rigs. It was in the most recent podcast. This was more of a just for fun and some info style test. Def agree with you though!

[@stevewells20]

@@802Garage Oh awesome, I was hoping it was something like that! CNCKitchen has fantastic physical test data, and I'd love to see more from Slant, especially if they publish the raw data!

[@johnsmith-i5j7i]

What infill % would you recommend for display models ie tanks cars etc?

[@philipzielinski]

Designing a part for compressive force, you'd want the orientation to have the infill stand as tubes rather than laterally across the seams. Testing that orientation would be useful.

[@joshuacheung6518]

I think the idea might not be to test against the strongest axis? Not sure.

[@david8694]

​@@joshuacheung6518 When designing a column, or anything that you know will experience compressive force, you want to load it perpendicular to the grain (in this case, perpendicular to the infill layers) because you know it will be stronger and therefore a more efficient part. Parallel to the grain is weak, aka the weak axis. This applies to timber as well. If you anticipate compressive force, you will want to know how strong your material is when loaded axially, perpendicular to the grain, because you can get away with much smaller member sizes since your part will be stronger. If you can design your compressive part in such a way that loads are applied perpendicular to the grain, your part will resist force more efficiently. Tension is the opposite. You want tensile members to be loaded parallel to the grain, or the layers. So while that may not be the idea of the video, it most definitely is in design which imo is the entire point of performing an experiment like this. If you know you are designing a structural member, you first want to know your loading which will then allow you to design your part for certain stresses. The experiment should be done accordingly because again, if you know you are designing a structural member, you are designing for efficiency, and therefore you want to know how much force your part can resist in compression against its strongest axis. You wouldn't choose a 12x12" column loaded on the weak axis when an 4x8" can support the same load combination on its strong axis, just as an example, and similarly, you'd want to save your PLA or ABS if you can get away with a smaller part to support the same load.

[@alias5281]

@@joshuacheung6518 Why not? Using the wrong axis doesn't give much useful data.

[@likeggs6465]

Thanks for the tip, I was just about to print something that will be used in a compression situation. The only thing I'm uncertain about now is which type of filiment to use, PLA-CF, PETG-CF or ABS.

[@SaHaRaSquad]

Imho this would be much more useful with more realistic configurations, like cubic and gyroid infills, fully enclosed walls, and different perimeter numbers.

[@slant3d]

What infill density do you typically aim for?

[@orangethesaberguy7638]

I find 15% cubic I fill works well for almost every project I have unless I need a completely solid part, which I use 100% rectilinear. Gyros was my go-to for a while but I find that cubic works well and doesn’t wobble my printer as much due to the straight lines as opposed to the wavy pattern of gyroid. Most of my projects aren’t under any substantial load though

[@sleepib]

Usually I use infill for cosmetics, just enough to support the top of the part, and add perimeters for strength.

[@notsaucemanofficial]

35-40% for triangles and 14-20 for gyroid

[@JonasE-l7q]

Please measure and caculate the strength per weight next! :)

[@QFrewin]

I'm normally around 15%, but I haven't done too much with different infill types.

[@ExplosiveOldMan]

How is the strenght in compression in the z printing axis? With printing in the proper orientation, shouldn't you try to align the greatest force perpendicular to the layers, so it doesn't fail in shear due to weak layer adhesion? It's an interesting video, but with proper design and orientation during printing, which you channel is mostly about, one hopefully shouldn't be in a position where the force is applied in line with the layers.

[@RobertCowanDIY]

I'm curious if the rise after 60% had anything to do with the change in the test setup. Switching from a wood base to arbor plates would most certainly make a substantial difference, since the wood is absorbing some of the force. Also, did you really break your setup, or did you just max out your load cell?

[@noahkatz9616]

It seems that the results might be pretty sensitive to how well the ram is centered on the part. Small differences in distance from the edges and corners will make big differences in the amount of local bending/buckling failure. A load-spreading plate that covers the entire top surface of the test item would address that.

[@archkittens]

loved seeing the wood break before the plastic towards the 70% infill

[@GyrosoftCo]

I think Gabe has a secret career in sports commentating.

[@AwestrikeFearofGods]

I like that you're compression testing the specimens in the transverse direction: 1) This type of data is rare, making it more interesting to me. 2) Extruded-section core (e.g. grid, honeycomb) is much stronger in axial compression, because it's naturally more resistant to buckling and shear failure when loaded along the cell axis. There is even more scatter in the data when compressing axially, since the initiation of buckling is sensitive to small differences in defects, loading direction, moon phase, and planetary alignment. The data wouldn't be as meaningful unless you were willing to slog through several specimens per density to gather statistics.

[@McRootbeer]

This was awesome! Hope to see more tests, especially gradient infill like what Stefan from CNC kitchen did. It'll be interesting to see how much filament might be saved by using it. Can't believe it hasn't been implemented in any slicer yet.

[@neonnerd1364]

I find myself using 60-80% infill more often than any other percentage. I also tend to print things that need that strength. For example a thread on shift knob for my stick shift daily driver which was printed at 70% infill. Been slamming gears hard on purpose to see if it'll break and so far after 6 months it's been just as solid as the day I put it on.

[@HappyJackington]

I would be really interested in seeing you test how wall thickness affects strength. In my anecdotal experience it ends up adding a decent amount to part strength for only a minor cost in print time. Regardless, thank you for doing this testing!

[@ahsanatabig6071]

3:25 prety solid MMXM demostration there. and classic god candle at the end

[@B0A2]

Audio quality was bad near the end of the video. Would love to see more infill pattern and the effect of pressing down on it from the side that was facing the camera

[@B0A2]

Should have watched to the end before commenting looking forward to future videos

[@andrewtinker7537]

If you want to know the strength of the higher infill blocks, there's a way, without having to buy a bigger load sensor. By resting a rigid bar of metal on two prism shaped fulcrums, one of which is on the press bed and the other on the load sensor, you can adjust the capacity of the sensor by putting the test block at different places along the bar. For example, if it's centered between the fulcrum on the bed and the fulcrum on the sensor, you're doubling the capacity of the sensor by splitting the load on it in half. If you put the test block at the 1/3rd point of the bar, you triple the capacity by splitting the load into thirds, etc.

[@richdobbs6595]

I use 100% infill on small parts, and typically its shear rather than compression loading.

[@atomicsmith]

Would have been interesting to also graph the print times to see if there is an optimal compromise.

[@oasntet]

And also actual weight, though this will correlate loosely with print time. That's going to be far more important than infill %, especially when comparing across infill types where infill % doesn't really mean the same thing.

[@AwestrikeFearofGods]

@@oasntet Yeah. It's strange that 20% doesn't always mean 20%.

[@StumblingBumblingIdiot]

I usually go for 25% gyroid with usable models and 5% or less for just statues/busts/etc. Now the really neat thing to see would be how much difference there is if you add walls....1,2,3,4,5,6,how ever many you want and see if it actually helps as much as I think it does.

[@tonyhartmann7630]

Awesome testing. More testing like this please. Also other infills like cube etc. Also other materials … like PA12, PA12-CF, PCTG and PP. Thanks a ton 😊

[@kyleheaser351boombaby]

Huge error here which completely voids this entire video. The downward pressure should not occur inside of the edges of the cube, but rather passed the edges of the cube. This would provide equal force on the entire surface area where pressure is being applied. On the fully filled cube, it is closer to being properly dispersed pressure, not quite perfect still, but closer. Had you had a metal pate of a proper thickness at the end of the cylinder that was wider than the cube this would be been a perfectly fine test.

[@davedixon2167]

This is a great start but I would personally need to see the same with the part laid down such that the infill is parallel to the bed - I make products that are intended to be stepped on, and I only go 10 to 20% infill, but my items also have things like vertical cylindrical holes and arched tunnels across the bottom, so the infill itself doesn't have to support the bulk of the force.

[@plasticlawnchair7197]

was thinking the same thing

[@BionicKing]

Do you have a blog or something I could follow? This seems like it would be a lot more convenient for me to process this kind of information in a text-based medium.

[@adamsvette]

THANK YOU. It doesn't matter if the numbers dont matter. its just fun to look at the numbers

[@chinook9785]

I really enjoy your infill testing videos. It's fascinating to see that what I've read about in my bachelor thesis can be observed in your tests as well. However, there's still a lot of room for improvement in your method to obtain reliable and useful data. I find it interesting how infill changes the cell size to alter relative density, compared to the literature, which typically modifies wall thickness to adjust relative density for lattice materials. Perhaps this is because most studies use powder bed fusion (PBF) or UV-resin printing for testing. Please be aware that when comparing different infill patterns, the strength of the pattern is often anisotropic, meaning it depends on the loading direction. I'm not referring to the anisotropy created by the layers. For reliable comparisons of different patterns, you need to test different orientations of the infill pattern. I understand that running so many tests can be challenging. If you want truly reliable data, it would be necessary to test multiple samples of each test sample to understand the spread of the results too, like CNC kitchen often does.

[@chinook9785]

The issue with this is that for low relative density, you have too few cells, leading to increased edge effects. This means that a significant portion of the load is borne by the outer walls, which can distort your results.

[@chinook9785]

I would not agree with your assessment that 100% infill behaves like a solid plastic part. It can be clearly seen that there is crack propagation along the diagonal, a result of the infill pattern. You are right in your assessment that the part does not fully shear off. I think this could be because the part reaches the densification region before it shears, meaning all the walls touch each other, resulting in an increase in stiffness.

[@chinook9785]

You should weigh your parts to verify their relative density. The produced parts realitve density can differ from the targeted relative density.

[@chinook9785]

Just that you know I have so much to critic on your test, that it can seam, i dont apreashiated them, i really do. I am so exicted about this topic. There is so much potential to improve infill and so much to learn. The Topic is so fascinating. The litrature about lattice structures is just so inacessable. There are so many papers about the topic and i have not found any Good books on the topic which summerizes the field. To recommend to you.

[@LearnEverythingAboutDesign]

@@chinook9785 And on these points comparing a 3d printed part, even with 100% infill can't be "just about the same dynamics as a solid plastic part". Even if you print 100% infill it isn't the same crystalline structure as a solid plastic part and the interior of the part still has a pattern to it at some angle, which we can see here from the failure. It will never behave like a solid part, but we don't generally make 3d printed parts solid anyways. You want to put a hydro jack on some plastic parts and say look how they fail, cool, and there is validity in trying to understand that. But you can't make scientific statements about "isotropic" materials and their performance when testing a single print on top of a block of wood. I think what would be a more valid test is to somehow validate layer line adhesion as that is generally the root failure of an FDM/FFF part. Or get some PETG raw bar stock and print some PETG parts and do the same test with a rigid base and fixture to apply an even force to the top of it if you want to make a claim about "as good as a solid part".

[@spencerlange125]

I would like to see another test, like one where you test the strength of PLA parts at various ambient air temperatures. We would be able to see what the temperature limits are for functional PLA parts. I don't ever use PLA for functional prints but it would still be nice to know.

[@nullstyle]

I'll fuck around with some scripting over the next week to see if I can pull structured data from the videos to put up on a website

[@polycrystallinecandy]

They already have structured data

[@nullstyle]

@@polycrystallinecandy thanks for the enlightening response 🙄

[@polycrystallinecandy]

@@nullstyle I'm just saying they intend to publish the results anyway (according to their other video where they discussed doing extensive filament testing)

[@nullstyle]

@@polycrystallinecandy If you remember, they asked for help with the effort. I'm just saying "Hey, I'll see if I can help with this effort in some way", by writing some code around what they are already publishing to automate some of the work in publishing something besides a KZbin channel. If they start posting a directory of JSON with contents good enough to drive a website, then sweet I'll use that but I think it's fair to say we're still in early stages of development and any work I can do on my own will save other people time. Please leave

[@polycrystallinecandy]

@@nullstyle use your time however you like bro, why are you getting so worked up? I was just pointing out they might soon be publishing it anyway.

[@deanrantala]

In all these tests, the strength is tested in the diagonal. Notice they always shear at a 45 degree angle? If you placed these "up and down" where the webbing is perpendicular in some way to the force being applied, you will see a huge increase in break resistance. Essentially making it so that the downward force is applied straight down to the wood -vs- cantilevering over each "row" below. Even better - turn those prints on end, so each square tube is straight up and down.

[@JohnR-fc7vr]

That was very imformative! Basically, if I don't reach a stress weight of 50 pounds throughout my structure, I should be good with 10% infill.

[@Nero_AZ]

Would love to see how much compressive force the infill pattern can take when its placed in parallel with the forces rather than perpendicular. Loved the video though👍

[@levoniust]

I hope you come back to this with more equipment. And doing a full test of infill percentages up to 70% and different types of infills and different plastics. Since you're already doing the work it would be nice if you could add in other things like relative humidity in the air while printing and testing as well as the temperature of the print the room and the testing environment. Good luck!

[@justryingmybest]

Thank you for your contributions!

[@FluorescentApe]

I like these kinds of tests, letting us know where the point of diminishing return is. I would also like to see similar tests but with a plate that crushes the whole cube instead of only the middle. Including the walls would have give a bit of a different result.

[@hjimenez26]

Loved this

[@slant3d]

Thanks

[@death-ot2oe]

I would like to know what different infill shapes would effect

[@Ataraxia_Atom]

I dont remember who it was but someone did a study like this and it was triangles that was the strongest pattern so i usually just use triangles and 25-35% infill whether i need it stronger or not

[@no-page]

In an earlier video, you said you use Ingeo 3D850 resin; why are these test samples 4043D?

[@ImSoMagic]

I'm watching this like it was a sport game haha good video!

[@rooicarus6729]

i want to see whether releasing the presure before it breaks and coming back later makes it break at lower pressures afterwards

[@jeffmcclain]

Would be nice to know the failure force divided by the sample weight of filament. I would expect that to be a curve with a local maxima...maybe around 30%?

[@FranklyPeetoons]

Another excellent and info-packed video for the idly curious. Maybe next time, just for the heck of it, we can see the weight of each print to provide the relative mass of the material used without having to do maths.

[@khoroshen]

I’m thinking in kg, but I noticed something is wrong with your conversion to pounds. The factor you are using is for pound-force, not for pound (mass) which was likely the intention…

[@nullstyle]

Very cool content, thanks for the work!

[@martinfitzpatrick4703]

If you want to get more consistent results you could try using a smaller cube or a bigger loading platen. Everything you do should be the same. Making the walls thinner would test more of the infil strength and less edge effects. If you are going to use the same setup position of the rod to the cube needs to be exactly the same each time.

[@dannymorrison4185]

How did you remove the 2 faces of the cube, did you remove with Fusion or did you cut?

[@MiG82au]

That's a load cell. Strain gauges measure non dimensional stretch/shrink/shear. Even if it uses a strain gauge internally with a known stiffness to derive load, it's an internal process that's not related to the strain of the part being tested.

[@calvinflager4457]

Thanks for the great information.

[@walterlang1848]

Nice video. I would love to see one testing how much torque they shafts of printed PLA can take.

[@OldFilmCameras]

Did you mention the print orientation? Id imagine that the results would be slightly different if the layers were parallel to the head of the press.

[@Liberty4Ever]

The layers are stacked in the direction of the camera's view, with zero top and bottom layers. The press is crushing across the layer lines. Rotating the cube around the viewing axis (the Z axis when printed) should have little impact on the test results.

[@3DThird]

How about you try increasing walls instead of infill? Usually the part gets most of it's strength from the parameters so it would be enlightening to see how the number of walls correlate to the part's strength

[@cloudcamdrones]

I increase walls instead of infill to make strong parts. Very strong parts I use 30% infill and 7 or 8 walls.

[@romainlmml3852]

I just have a question, what is the matherial used ?

[@potpie7644]

Can you add each sample's mass so we can calculate the specific strength of each specimen?

[@kianfernihough2750]

wouldt the lack of side walls printed make a difference, you had them open to see the infill but if you had walls wouldnt that increase the yeild strength of the parts as there bi latterally supported and resisted from "squishing" outwards

[@tomsko863]

1:32 - Good video. Thanks for doing this. Just a suggestion, please give a little bit of a primer when you start your video so people can follow along. I know what's happening from experience working with graphs like this and it still took me a minute staring at it to make head and tails of where you were pulling number from. A graph title, units on the axis, and labels would be a start. Or just circle the freaking -0.234 kN in the corner once so people know what to look at. You can see from the comments below, not everyone watches all of your update videos so they don't understand the scope of what you're doing. That should also be in the intro.

[@narancs5]

Easy trick for next time. Put a strong solid bar under the tested object and the sensor under one end of the bar. You can double the limit of the sensor this way. Not that accurate obviously but provides a good estimate.

[@Deus-Ex-Blastlord]

There's the most random minecraft skeleton hurt sound in the video at 4:07 to 4:09. Idk if that was an accident or on purpose.

[@ttvfabian1157]

Did you use PLA filament?

[@Liberty4Ever]

That's some strong PLA! I gotta pick up a three pack at Tangled Filament (which isn't tangled at all). Great job with the real time strain gauge graph. It'd be great if you had a tiny hydraulic pump so the pressure is applied more uniformly than pumping a bottle jack. I'd love to see a test with other infill patterns, particularly gyroid which I've been using under the assumption that it would have more isotropic properties and wouldn't catastrophically fail on the diagonal as the grid infill does. PrusaSlicer allows the Fill Angle to be changed. For example, the grid infill shown in the video is the default 45 degrees. If you knew the applied force was vertical as in the test, make the grid infill at 90 degrees so there are short columns taking the load with horizontal cross bars to prevent buckling. It would be very interesting to see a comparison of grid infill at 45 and 90 degrees for these vertical loads. And... get a bigger strain gauge. 🙂

[@MattManuel]

I would like to see some more ratios for the comparisons. Like "force/gram of material" and "force/print time" to better understand the tradeoffs. Future video?

[@orphax1925]

to properly test the infill you should move the walls further apart, in these tests the walls are helping the part

[@seanwoods647]

I never used to consider infill as contributing much to strength. This was eye opening.

[@marcosmoura911]

It's obvious

[@enesozbey8346]

Can you use triangles instead of squares

[@rainzer0]

Its missing a wall layer which would greatly increase the strength. Very insightful video. Thanks for the hardwork

[@antonmaxx]

Excellent and thank you!

[@saltwaterrook4638]

Ok, I need this tested for every infill pattern!

[@slant3d]

already done

[@youtubehandlesux]

@@slant3d At every infill percentage? You didn't even test 80%-100%.

[@SnappyDrift]

I’d like to see now different material types!

[@yargienimanuel]

Hi... may i know the outlayer width? 1mm?

[@topwaifu2104]

Would be interesting to also check a point at which first irreversible damage appears and max elastic deformation distance.

[@sfkjbg]

This was an awesome video

[@justaguy4788]

This is perpendicular to the infill, Have you tested parallel to the infill also?

[@JohnnyC4G]

Nice Video pls more.. hello from Germany... We say in Germany... Das Video war sehr interessant und informativ danke.....

[@MarkPryor1]

45° angle not great for 0° load vector. Rotating the infill by 45° and reducing the 90° infill by 50% that of the 0° deg infill will give you much higher strength and stiffness. Also, the load distribution at the ends is essential. Need to spread the load first so more material is needed at the ends.

[@RWB123]

Another good video 📹 👏

[@softimad]

thank you for this interesting test, it's very helpful in choosing the right infill percentage. I'm thinking now if those results still the same if you change the wooden platform with a very regid one, i think the wood absorbe a litte bit of the applied forces. because i add sometimes a 3mm sheet of wood between the 3d printed part in the metalic surface of contact to reduce the applied forces.

[@romainlmml3852]

what is the material used ?

[@3sotErik]

I think it is possible to get more usable information from a test like this if you change just a couple things. Most of us are not printing functional parts with the intent of it only becoming useless only after critical failure. I think the real question is, under what strain does it go from elasticity to plasticity? Stop the test at different pressures and measure if it has returned to its original size. Then continue to test and see at what point it is permanently deformed by 10 or 20%

[@mpart_woodlathe-stuff]

How about same parameters but with a range of infill wall thicknesses ?

[@Bastelsaal]

Thanks for the video. Now I regret printing some important parts using only 20% infill 😀

[@mikebertolini120]

You should try tò do also testing with organic shaped infill

[@podschi22]

Great stuff! I’m sure you did it but it would be great for future videos to show an average of multiple readings per infill. Having only one reading doesn’t eliminate any false readings (outliers) that you might have in there

[@Liberty4Ever]

I generally agree that more data is more better, however, the graph showing the strength of each 10% increase in infill density was very clean so I believe the test methodology was sound and this simple test produced good data.

[@AwestrikeFearofGods]

You're right. There's definitely scatter in the data, as you might imagine from the chaotic initiation of buckling and shear failure modes. The only problem with multiple test specimens per infill is the requirement for multiple times as much testing. He's a busy guy. Anyway, the strength/density relationship is strong enough to see a pattern, regardless of scatter. It's typical of hollow core materials to have compressive strength that is roughly a power function of density, with an exponent somewhere above 1. In other words, doubling density increases strength by more than double. From this data, the exponent works out to range from 1.24 and 1.84 based on consecutive data points. It's all over the place from data scatter.

[@Quinten0508]

Interested in seeing what the best infill % would be in terms of kN per gram of infill used!

[@dva_kompota]

Please do gyroid which is supposed to be equally strong in all directions

[@anitexs373]

Why did you not test with Triangular infil? Triangles is the strongest geometric shape and should be able to handle more with less?

[@Felipe-tw1wg]

More infill, more strength... yeah, that is to be expected. But what about efficiency? It would have been awesome to see a mass to strength ratio!

[@thebluemarker9209]

Absolutely amazing video! Can you also provide the weight of each block so we can know which precentage is the best in terms of lightest weight?

[@navb0tactual]

This

[@victorrzhang]

If you can do the test in a different orientation and compare that would help tons

[@galinrashkov7645]

I think the reason all of the test samples failed diagonally was the fact that the steel rod pushing them wasn't leveled at all and pushing much more to the left.

[@-zectarius-5265]

Amazing vid! I'll like to see others partner like triangles or others shapes. I wounder how the delta N Max would be between different shapes

[@802Garage]

Love this type of video. I think it will do really well for your channel!

[@coreyfro]

Normalize by weight and then balance infill and walls

[@Aptidude]

Insightful… surprisingly compelling haha

[@slant3d]

Thanks

[@How_Interesting.]

Could you please have metric conversions atleast on the screen somewhere too I don''t know pounds and I had too keep flisking off the video to understand what the data meant.

[@razvanduta3563]

From what I see, the parts are loaded in the wrong direction; if you would try axially (same direction with the infills) you would get very different results.

[@calebthibodeaux4422]

Axis labels on the graph would be appreciated

[@How_Interesting.]

While watching the tests it seemed like you were a sports commentator going over practically the same thing over and over again. This whole video would have been better as a three miniut slug as you discussed all the results while watching maybe one or two of the specimens being broken in the background. INCLUDE METRIC CONVERSIONS

[@556todd]

I’d like to see the same test done but with shell thickness

[@TeddyBear-lb9ho]

What's a tough filament with decent flexibility?

[@ullumski]

Weighing each cube would be helpful to better grasp the relation of material spent to the resulting compression strength.

[@NoName-zg2te]

assumptions: counting pixels, wall width is about 1mm-ish, calculations simplified to volmue instead of mass, but whatever. (every single calculation below is calculated on pretty inacurate numbers, so take it with a lil grain of salt :P) wall: (30^2-28^2)*30 = 116*30 = 3480mm^3 inside: 28^2*30 = 23520mm^3 10%: 0.2 kN - 3480 + 23520*10% = 5832mm^3 - 0.05747N/mm^3 of filament; 20%: 0.5 kN - 3480 + 23520*20% = 8184mm^3 - 0.06109N/mm^3 of filament; 30%: 1.0 kN - 3480 + 23520*30% = 10536mm^3 - 0.09491N/mm^3 of filament; 40%: 1.7 kN - 3480 + 23520*40% = 12888mm^3 - 0.13190N/mm^3 of filament; 50%: 2.5 kN - 3480 + 23520*50% = 15240mm^3 - 0.16404N/mm^3 of filament; 60%: 3.4 kN - 3480 + 23520*60% = 17592mm^3 - 0.19326N/mm^3 of filament; so for strength/mass ratio it looks like 10-20% is barely a difference, 20-60% it's pretty much linear increase and above that, if we assume its still linear, at 70% infill, it should break at around 4.5 kN, so i'd guess that somewhere between 60 and 70% it stops behaving like grid infill and becomes more like a rigid wall :D